Xingolak
Novel visualizer for translations to/from Basque language
Last verified:
What is Xingolak?
Xingolak is a web‑based translation‑alignment visualizer specialized for Basque (Euskara), with support for English, Spanish, and French as source and target languages. It allows users to input a sentence or short text and receive a side‑by‑side display of the Basque translation alongside its source‑language counterpart, highlighting how individual words and phrases map between the two.
The tool combines machine translation (via batua.eus) with natural‑language processing (Stanza) to show structural and syntactic links. An LLM then processes the raw NLP output into human‑readable explanations of each word’s role, such as case marking, tense, or agreement, so learners can see not just what was translated but why certain forms appear.
Key users include Basque learners, linguists, and language teachers who want to explore how Basque agglutinative morphology and ergative‑absolutive alignment differ from Romance and English structures. The visualizations help users spot mismatches between word order, morpheme boundaries, and case usage, making it easier to internalize less‑intuitive Basque patterns while comparing them to familiar languages.
Xingolak pricing
Pricing model: Freemium
The service currently offers a free tier with a hard limit of ten translations, reflecting low‑resource backend usage and constrained machine‑translation quotas. There are no visible paid plans or subscription tiers on the site; the author notes that the project is constrained by token limits and infrastructure costs, so expanded access or commercial‑scale usage is not yet available. Any future paid tiers would likely aim to increase the number of allowed translations and reduce latency by upgrading the hosting and API infrastructure.
Xingolak pros
- Specialized for Basque linguistics and pedagogy
- Interactive alignment visualization between source and target words
- Supports four main input languages: Basque, English, Spanish, French
- Integrates high‑quality Basque‑owned machine translation (batua.eus)
- Uses Stanza for precise morpho‑syntactic analysis
- LLM‑enhanced explanations of grammatical roles under each word
- Helps users see agglutinative morphology step by step
- Highlights case, tense, and agreement markers visually
- Sentence‑by‑sentence layout that keeps structure clear
- Allows clicking on individual words to inspect details and explanations
- Shows when translation errors occur and points them out explicitly
- Designed with feedback from language learners and linguists in mind
- Useful for self‑study, lesson preparation, and classroom demonstrations
- Open‑web interface that requires no installation
- Encourages community feedback and iterative improvement by the creator
Xingolak cons
- Limited to only ten translations per session or billing cycle due to backend constraints
- Translation latency of 10–30 seconds per submission
- Reliant on external services (batua.eus, Stanza, LLM API) so availability may vary
- May produce occasional mistranslations, especially for idiomatic or rare words
- No guaranteed preservation of user inputs or history across sessions
- Limited length support; only short sentences or brief passages are practical
- No advanced customization options for visual style or dependency layout
- Currently focused almost entirely on Basque; support for other minority languages is not featured
Frequently asked questions about Xingolak
What languages does Xingolak support?
Xingolak supports Basque (Euskara) as the primary structured language, with English, Spanish, and French as both source and target options. Users can translate from any of those three into Basque or from Basque into any of them, and the alignment visualization will show how each word maps across the selected languages.
Is Xingolak only for Basque learners?
While the tool is especially useful for Basque learners, it is also designed for linguists, philologists, and language teachers who want fine‑grained visualizations of Basque morphology and syntax. Native speakers and advanced users can leverage it to contrast Basque structures with Romance or English equivalents and to study translation errors or edge cases.
How does the word‑alignment visualization work?
After you submit a sentence, the tool runs the text through a Basque‑focused machine translator and then processes it with Stanza to identify parts of speech, lemmas, and morphological features. An LLM then converts that raw output into human‑readable explanations, which are displayed beneath each word, while lines or highlights show how source‑language words map to their Basque counterparts and vice versa.
Why does each translation take 10–30 seconds?
The delay comes from the chained pipeline: the Basque machine‑translation service must generate the initial translation, then Stanza parses the text, and finally an LLM reformulates the technical annotations into clear explanations. The long processing time is most noticeable on slower or resource‑limited hosting, which the author explicitly notes as a current bottleneck.
Does Xingolak correct translation errors automatically?
The tool does not yet fully auto‑correct mistranslations, but it can flag and explain when the LLM detects a mismatch between meaning and structure. The creator has mentioned plans to have the LLM automatically submit translation corrections in upcoming versions, which would tighten the loop between analysis and improved output.
Can I use Xingolak for long texts or documents?
Xingolak is currently optimized for short sentences and short passages rather than full documents. The interface and backend are built around individual sentence submissions, and trying to submit very long texts may either time out or produce unclear visualizations due to layout constraints and the ten‑translation cap.
Is my input saved or stored by Xingolak?
The site does not advertise any explicit user accounts or persistent storage, so inputs are generally treated as transient queries tied to the current session. There is no indication that sentences or translations are logged long term or tied to user identities, but the project uses third‑party APIs whose own data‑handling policies should be reviewed separately.
How accurate are the grammatical explanations?
The grammatical explanations are based on Stanza’s NLP pipeline and then smoothed by an LLM, so they are generally accurate for standard Basque forms. However, because the tool is still experimental, some edge cases or dialectal constructions may be mislabeled or oversimplified, and the creator explicitly invites feedback from linguists and learners to refine the analyses.
Can I share the visualizations or export them?
The website does not currently highlight built‑in export or sharing features such as downloadable images, PDFs, or embeddable links. Users who want to share outputs would likely need to take screenshots or manually copy the text and annotations and then distribute them externally, since the tool appears to focus on live interaction rather than offline sharing.
Will there be more languages added in the future?
The author has framed the project as an experimental visualization tool centered on Basque, and there is no indication of immediate plans to expand to other minority or non‑Basque languages. Any future language additions would depend on securing additional translation and NLP resources, and the current infrastructure is explicitly described as limited and tailored to the existing four‑language setup.